Skip to main content
Managed Deep Agents is simpler way to create powerful agents. It uses a powerful agent harness with opinionated infrastructure, allowing you to focus on the your business logic. Managed Deep Agents consist of three layers:
  • Business logic - this is what you are in charge of
  • Agent Harness - we use the Deep Agents harness
  • Opinionated Infrastructure - this is what managed Deep Agents adds over the open source Deep Agents Harness
The opinionated infrastructure consists of several pieces: Runtime: We use LangSmith Agent Server to run agents in a durable, fault tolerant manner Sandboxes: We us LangSmith Sandboxes so agents can write and execute untrusted code in a secure manner Evals: We make it easy to author evals in Harbor format, so you can properly test agents UX: We have a “channels” abstraction to allow you to easily bring your agent to platforms where your users live. Memory: We have opinionated memory so that your agents can remember interactions. Context Management: We use LangSmith Context Hub to manage agent instructions and skills so that you can modify them in a UI and have those changes take affect without having to redeploy.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.
Edit this page on GitHub or file an issue.